SmolLM: The Tiniest Local AI Model
Hugging Face's SmolLM proves that AI can run on nearly anything. SmolLM2 360M loads in roughly half a gigabyte and needs only 1GB of RAM, the lowest requirement in our entire database. It is also the fastest model we track by speed score. Quality is limited at this size, so think of SmolLM as a specialist: text classification, simple extraction, short completions, and basic chat on devices that cannot fit anything larger. For older iPhones, entry-level Macs, and embedded projects, it is the minimum viable AI.
- developer
- Hugging Face
- local models
- 1
- size range
- 0.36B-0.36B
- RAM range
- 1-1 GB
All SmolLM models
1 local models| Model | Size | Quant | VRAM | Min RAM | Best For | Quality | Ollama |
|---|---|---|---|---|---|---|---|
| SmolLM2 360M | 0.36B | Q4_K_M | 0.5 GB | 1 GB | Chat, Embedded | 38 |
Key features
- Tiny 360M-parameter model from Hugging Face
- Loads in about 0.5GB, needs just 1GB of RAM
- Highest speed score of any model in our database
- Runs on old Macs, iPhones, and embedded or IoT devices
- Handles classification, extraction, and short text tasks
- One-line setup: ollama run smollm2:360m
Device compatibility
Which SmolLM models can run on each device class, based on minimum RAM requirements.
| Model | iPhone | Air | Pro | Studio | Mini | |
|---|---|---|---|---|---|---|
| SmolLM2 360M (0.36B) | Excellent | Excellent | Excellent | Excellent | Excellent | Excellent |
RAM requirements
Largest dense Q4 model that fits each memory tier
Get told when a better model fits your MacBook Air M5 16 GB
Get one email when a new open-weight model is a better fit for this machine.
For: MacBook Air M5 16 GB
You'll also get the weekly ModelFit email. By signing up you agree to our Privacy Policy. Unsubscribe from either email anytime, in one click.
Config reports
No speed test run for this family yet. Run the speed test: npx @modelfit/cli@latest bench
Share your config
Frequently asked questions
What tasks suit a sub-1B model like SmolLM?
Short, narrow tasks: text classification, keyword and field extraction, autocomplete-style completions, and simple one-turn Q&A. Multi-step reasoning, long conversations, and coding are beyond a 360M model. Step up to a 2-4B model for those.
Which devices can run SmolLM?
Almost all of them. With a 0.5GB load and a 1GB RAM floor, SmolLM2 360M runs on every Apple Silicon Mac, older Intel Macs, iPhones, and small single-board computers used in IoT projects.
Is SmolLM useful or just a demo?
It is genuinely useful for specific jobs: classification, extraction, and basic chat on devices that cannot run anything larger. It also works as a zero-cost test bed before wiring a bigger model into the same pipeline.
What Ollama command runs SmolLM?
Run `ollama run smollm2:360m`. The download is small and the model starts almost instantly, even on machines with very little free memory.
